Jun 21, 2012


A couple weeks ago, Borepatch put up a piece on what happens when you fake fudge fill in supplement estimate your data points. In particular, he mentions urban bias.

It's been warm around here today, and I glanced at the Wunderground for the local weather stations, finding this:

Average temp across those stations? 90.4F.

Out in the boonies and on the hills, the temps are ringing in the mid-to-upper 80s. The second one there, "Neighborhood", at 95? That's downtown. It's in a tree-lined shady residential area. Two lane streets, lots of shade, but lots of houses.

Nix a few of those outlying areas - say the lowest three temps - and the average is now 91.2.

But what if we "extrapolate" our data from "Neighborhood" and fill those in for the three lowest? The new average is now 92.

Urban bias is real. Picked data sets aren't real data.

1 comment:

Old NFO said...

Very true... sigh